toone

Overview

How Toone works

Step-by-step guide to Toone on your Mac: organizations, agents, routines, schedules, approval gates and integrations, from first setup to reviewed runs.

Product guide · Overview

Start here

OverviewGetting started

Getting started

Install and connectCreate a projectCreate an agentCreate a routineRun and debugImport from ExploreShare on Explore

Understand Toone

Core concepts

Features

Overview

Experience

Agent SpotlightWorkspace windowsZen modeProject switcherProject Explorer

Workflows

Browser sessionsRoutinesAgent orchestrationExploreCalendarBackground threads

Control and safety

Project HistorySafe Mode

Connect

Live ShareVoice and meetingsMobile companionMCPs and integrationsAI providers
Toone product guide

What it is and who it's for

Toone is the AI workspace for your agentic workflows: a macOS app where specialist agents run your workflows and routines under your control. It is for anyone with recurring, multi-step work, such as a weekly client report, a literature scan or a launch checklist. You talk the work through with an agent and set it up once as a routine, then get an output you can check, run it again and share it. You need a Mac and your own Claude Code or Codex. While a routine runs, you can watch each step, pause it, edit it and review the result before you use it.

One workspace: chat on the left, a working panel such as Routines in the middle, and Project Explorer on the right.

You can begin with one agent and one useful task. As the project grows, agents can specialize, routines can call other routines, and work can move between roles without turning into one untraceable conversation.

Not sure where to start? Browse the routines and bundles other people have shared in Explore, then open one in Toone. Import a routine from Explore shows each step.

The operating loop

Most work in Toone follows the same loop:

  1. Define your agents. Give one or more agents a clear purpose, explain what each is responsible for, and describe how they should approach the work.
  2. Shape the process together. Talk through repeatable work with an agent, then turn that conversation into a routine written in natural language.
  3. Run and observe it. Follow the work while it happens and inspect where it slows down or fails.
  4. Keep the result. Outputs, project files, and useful context remain available for the next run.
Define agents, shape a routine with them, run and observe it, and keep the result for the next run.

This makes Toone useful both for direct work with an agent and for processes that should run the same way again.

What lives in a project

A project is the boundary around your work. It can contain specialized agents, repeatable routines, connected tools, readable files, and accumulated knowledge. Those parts remain related instead of being scattered across unrelated chats.

A project keeps its organization beside its files: skills, departments, memory and connections.

The project begins as simply as you need. You do not need a large agent team or a finished automation plan before starting.

Start with a useful result

The fastest way to understand Toone is to complete one small loop: connect an AI provider, create a project, add an agent, describe a routine, and run it.

Setup opens on a choice of route. Express covers the essentials and is the recommended start.

Start the guided setup.

Explore at your own pace

If you prefer to understand the model first, read Core concepts. If you already know the basics, browse the feature overview.

Part of the draft's graph: each step shows what it takes in (↓) and puts out (↑), and step 2 calls a sub-routine.